parts detected by logical cameras should have unique TF frames

[Deanna Hood]

currently if two logical cameras see the same part, the TF tree is continually re-configuring because both publish transforms with the same child but different parents. this is inefficient at best and risky at worst (tfoote was surprised it was working as-is).

we'll need to update how we publish the world->agv1 transforms because now it needs to be world->agv1_as_seen_by_logical_cameraN

[Deanna Hood]

see https://bitbucket.org/osrf/ariac/pull-requests/50

[Deanna Hood]

Addressed in https://bitbucket.org/osrf/ariac/pull-requests/50

Users will need to specify prefixed names of TF frame IDs, i.e. specify what sensor is providing the transform data. e.g. rosrun tf tf_echo world logical_camera_2_kit_tray_1_frame. tutorial updated.